Anomaly in mpn_sqrtrem and mpn_rootrem

[Torbjörn Granlund]

  > Surely the most important k is 3 (assuming 2 is always handled by the
  > sqrt code).
  
  Elaborate? Are you thinking of any particular algorithm using cube
  roots, or just the general observation that smaller numbers tend to be
  more common?
  
The latter; I think u^(1/3) will tend to be more often computed than
u^(1/4711).

-- 
Torbjörn
Please encrypt, key id 0xC8601622